Ultima Company Macnica (headquarters: Yokohama City, Kanagawa Prefecture; President: Kiyoshi Nakashima; hereinafter "Macnica"), which imports, sells, and develops ALTIMA Company and network equipment, Digital Media Professionals Inc. (headquarters: Nakano-ku, Tokyo; President: Chiyoda-ku, Tokyo; President: Hiraga Toshiki; hereinafter "Morpho"), which engages in research and development, product development, and licensing of image processing technology, are pleased to announce that they have begun a collaboration on AI (artificial intelligence)/deep learning technologies using ^Intel® FPGAs.
As the first step, DMP will package an ^Intel® FPGA-based deep learning inference environment into a platform that is scheduled to be sold by Macnica ALTIMA Company during fiscal year 2017.

Artificial intelligence/deep learning has been attracting a lot of attention recently, and its application in a wide range of fields is being considered, with an increasing number of cases in which it is being adopted by end users such as manufacturing, wholesale/retail, transportation, and services. However, there are few effective solutions on the market for deep learning inference environments that can achieve the flexibility, real-time performance, high processing power, and low power consumption required for edge computing and some fog computing. , has become one of the impediments to the introduction of deep learning in the embedded field.
Through this partnership, the platform that the three companies will start providing will include Macnica 's deep FPGA technology, DMP's advanced processor technology and hardware optimization technology, and Morpho's neural network design and model optimization. It is a fusion of each technology. We are confident that this will accelerate the introduction of deep learning inference environments in real business in the edge computing and fog computing fields.
By introducing this platform, you can not only develop and evaluate based on trained models of real object recognition DNN (Deep Neural Net), but also programmably implement trained models prepared by users. It is possible to quickly develop/evaluate an inference environment and apply it to embedded devices in mass production.

Tatsuo Yamamoto, President and CEO of DMP, commented as follows.
“Through this alliance, we believe that it will be possible to build an effective system that fully utilizes edge AI at low cost and in a short period of time, which has been difficult to achieve in the past. We expect that it will enable us to quickly provide flexible and up-to-date solutions in the field, and that it will also lead to the development of new applications using AI.”

Morpho's President Hiraga commented:
"Through this business alliance, we will cooperate with each other to spread the fusion technology of the three companies in both hardware and software by leveraging their respective strengths in the AI field and their highly specialized technical capabilities, into actual business not only in Japan but also in the global market."

Yusuke Kobayashi, Company President of Macnica ALTIMA Company, commented: “We are confident that the outcome of this partnership will be a realistic edge AI solution that takes advantage of the acceleration characteristics of FPGAs.We will continue to provide AI solutions through the collaboration of the three companies and aim to become a market leader. We will do our best to revitalize it.”

<Example of object detection by Morpho's AI (artificial intelligence) / deep learning>

【Product Summary】

Product name: Deep Learning Inference Platform for ^Intel®​ ​^Arria® 10 GX

Contents: Mpression Magnes board
*Magnes board overview link: https://www.macnica.co.jp/business/semiconductor/macnica_products/boards/125973/
FPGA design including DMP's AI processor IP "ZIA™ DV700"
* “ZIA DV700” overview link: http://www.dmprof.com/jp/product/deeplearning/zia_dv700.html
Object recognition DNN (Deep Neural Net) trained model

Supported Frameworks: Caffe, TensorFlow, Keras

Supported Networks: Supports major neural networks such as CNN, RNN, LSTM, GoogLeNet

Target applications: General vision applications

• Driving safety support system
• In-vehicle infotainment
• Surveillance camera
• network video recorder
• AR/VR
• machine vision
• medical camera

This product will be exhibited at the DMP booth at the International Imaging Equipment Exhibition 2017, which will be held at Pacifico Yokohama from December 6 (Wednesday) to 8 (Friday), 2017.
■ Exhibition name: International Imaging Equipment Exhibition 2017 https://www.adcom-media.co.jp/ite/
■ Schedule: December 6 (Wednesday) to 8 (Friday), 2017
■ Place: Pacifico Yokohama Exhibition Hall D DMP: Booth No.76

*Company names and product names mentioned in the text are trademarks or registered trademarks of their respective companies.